Evaporative cooling of continuously drawn glass fibers by water sprays

Evaporative cooling of continuously drawn glass fibers by water sprays

International Journal of Heat and Mass Transfer 43 (2000) 777±790 www.elsevier.com/locate/ijhmt Evaporative cooling of continuously drawn glass ®ber...

428KB Sizes 3 Downloads 93 Views

International Journal of Heat and Mass Transfer 43 (2000) 777±790

www.elsevier.com/locate/ijhmt

Evaporative cooling of continuously drawn glass ®bers by water sprays Matthew Sweetland, John H. Lienhard V* W.M. Rohsenow Heat and Mass Transfer Laboratory, Department of Mechanical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Rm 3-162, Cambridge, MA 02139-4307, USA Received 24 August 1998; received in revised form 6 May 1999

Abstract This paper examines the e€ect of the water sprays commonly used to cool freshly drawn glass ®bers. A model has been developed using a Karman±Pohlhausen treatment of the velocity and thermal boundary layers and accounting for the evaporation of an entrained water spray. Solutions of the model equations have been calculated, and the e€ect of changing various process parameters is studied. Variations in Sauter mean diameter, spray density, and spray placement along the ®ber are considered, as well as the e€ect of ®ber diameter and drawing speed. Fiber temperature pro®les for di€erent values of the process variables are presented. # 1999 Elsevier Science Ltd. All rights reserved. Keywords: Glass ®ber; Spray cooling; Heat transfer; Continuously moving surface; Evaporative cooling; Integral method

1. Introduction Glass ®bers are used for reinforcement of products ranging from structural plastics to fabrics. Such ®bers are made by drawing molten glass through an array of small diameter bushings. The glass solidi®es within a few centimeters of the bushing and is then pulled a distance of 1±2 m prior to coating with various binders or sizing compounds. Typical ®ber diameters are 5±30 mm with drawing speeds varying from 15±90 m/s and array sizes ranging from several hundred to several thousand ®bers. This process has been in use for several decades [1]. The cooling of the ®bers from the glass extrusion temperature (1500 K) to a temperature where the coatings can be applied (below 365 K) is a rate limiting step. The high ®ber speeds and short drawing lengths

* Corresponding author.

require a temperature drop of over 1100 K in roughly 25 ms. In order to enhance the heat removal from the ®bers, the ®ber bundle is sprayed with water from atomizing nozzles. Typical Sauter mean droplet diameters are of the order of 15±20 mm and droplet number densities may range from a few hundred per cm3 to a few thousand per cm3, depending upon nozzle and air¯ow conditions. At the ®ber temperatures involved, water cannot wet the ®ber surface without an initial quenching process. Moreover, at typical number densities, droplet speeds, and ®ber speeds, direct collisions of droplets with the ®bers are infrequent and will involve only a small percentage of the ®ber surface. Thus, for most of the ®ber, cooling enhancement is evidently due primarily to evaporative e€ects in the air adjacent to the ®bers. At present, no quantitative method is available for calculating the e€ect of the water spray on the cooling rate of the ®bers. Production stations have been designed through a trial and error process in nozzle

0017-9310/00/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 0 1 7 - 9 3 1 0 ( 9 9 ) 0 0 1 8 0 - 5

778

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

Nomenclature a a1, a2, a3 A1 B B1 C1 cp cpf D0 D hfg k kf kevap m_ 000 n0 n p Pr Q q_e

®ber radius (m) b-series expansion coecients for small a kevap curve-®t coecient (m2/s K2) …krcp †=…kf rf cpf † kevap curve-®t coecient (m2/s K) kevap curve-®t coecient (m2/s) air speci®c heat capacity (J/kg K) ®ber speci®c heat capacity (J/kg K) initial droplet diameter (m) droplet diameter (m) latent heat of vaporization (J/kg) air thermal conductivity (W/m K) ®ber thermal conductivity (W/m K) evaporation coecient (m2/s) local volumetric evaporation rate (kg/m3 s) environmental spray density (drops/m3) boundary layer spray density (drops/m3) boundary layer temperature pro®le coecient Prandtl number convective heat ¯ow from ®ber (W) volumetric heating due to vaporization (W/m3)

placement and operating conditions. Determining the e€ects of changing process parameters, such as spray size and density, is only possible through direct experimentation on production systems, which is both time consuming and expensive. A model is needed to calculate the e€ects of evaporative cooling on the manufacturing process. Such a model can be used for improvement of the existing process and as a design tool for developing new processes and products. The basic problem of boundary layer growth and heat transfer from continuous cylinders moving through a still ¯uid has been studied quite extensively. Early work by Seban and Bond [2] and Glauert and Lighthill [3] examined stationary cylinders in a moving ¯uid. Sakiadis [4] showed that the case of a moving cylinder in stationary ¯uid leads to di€erent ¯ow and temperature ®elds; he developed a momentum integral solution for the boundary layer ¯ow around a moving cylinder using a logarithmic velocity pro®le. These integral solutions have been used in many subsequent studies [5±8], and exact solutions of the boundary layer equations by Sayles [9] lie within 9% of the integral solutions. Convection from isothermal ®bers was studied using integral results by Bourne and Elliston [5], assuming constant ¯uid properties; exact solutions

q t T Te Tf u U v x X y a b d dT Z Y Yf m n r rf rw

boundary layer temperature pro®le coecient time (s) boundary layer air temperature (K) environmental air temperature (K) ®ber temperature (K) axial boundary layer velocity (m/s) ®ber velocity (m/s) radial boundary layer velocity (m/s) axial distance from bushing plate (m) xn=Ua2 radial distance from ®ber surface (m) velocity pro®le parameter temperature pro®le parameter momentum boundary layer thickness (m) thermal boundary layer thickness (m) dimensionless volumetric energy ratio …ˆ …rcp †=…rf cpf †† boundary layer air temperature defect (K) ®ber temperature defect (K) dynamic viscosity (kg/m s) kinematic viscosity (m2/s) air density (kg/m3) ®ber density (kg/m3) water density (kg/m3)

were obtained by Karnis- and Pechoc- [8] and were found to lie within 10% of the integral results. In addition, Beese and Gersten [10] have provided perturbation solutions for isothermal ®bers that are valid in the limit of large ®ber diameter. Of greatest relevance to our case are those studies that account for the ®nite heat capacity and axial temperature drop of the ®bers. Bourne and Dixon [6] used integral solutions for the cooling ®ber, neglecting conduction resistance within the ®ber and using averaged values of the ¯uid properties. Their results showed good agreement with experimental measurements by Arridge and Prior [11] and Alderson et al. [12] on 15± 50 mm diameter on glass ®bers. They showed that attention to the averaging of temperature-dependent air properties was quite important. Chida and Katto [7] subsequently amended the Bourne and Dixon theory to account for radial heat conduction in the ®bers; for conditions typical of glass ®bers, radial conduction proves to be negligible. Chida and Katto also developed scaling parameters for the problem, which were shown to collapse the above mentioned experiments. Maddison and McMillan [13] reported further experiments on ®ber cooling which were in general agreement with the results of Arridge and Prior. Kang et al.

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

[14] performed experiments on large diameter aluminum cylinders drawn through air and water; they found good agreement with the scaled theoretical results of Chida and Katto. Richelle et al. [15] provided exact solutions of the boundary layer equations for moving ®bers. Roy Chaudhury and Jaluria [16] performed numerical integrations of the full elliptic equations and found agreement with the experimental studies that was similar to that of Chida and Katto. Together, these studies provide strong validation of the integral theory originally developed by Bourne and Dixon. These studies focus on laminar ¯ow in the boundary layers; at present, little is known about the stability of these boundary layers. Several of the above studies [5,8,10,15], and others [17,18], have also produced results for the Nusselt number of ®bers moving through dry air. We are aware of no work addressing spray cooling of glass ®bers, although a large body of work deals with other problems in spray cooling. Such studies include cooling of atomized ¯uids [19], cooling of large ¯at surfaces [20,21] through direct spray impingement on the surface, and cooling of air by water spray [22,23]. A recent study by Buckingham and HajiSheikh [24] considered spray cooling of large metal cylinders. Of particular relevance is their identi®cation of a convection-dominated cooling regime in which water spray was found to evaporate within the boundary layer of a high temperature surface, without actually impinging upon it. This process is quite similar to the one modeled here for glass ®bers. In this paper, we use the Karman±Pohlhausen integral method to develop a set of equations that describe the growth of the ®ber thermal boundary layer with droplet evaporation occurring in the boundary layer. In the limit of no spray, our approach is identical to the Bourne and Dixon theory [6], although we have not assumed constant ¯uid properties along the length of the ®ber. The boundary layer evaporation term is evaluated using a single droplet model, and the resulting equations are used to calculate ®ber temperature pro®les for various spray and ®ber conditions. The ®ber is assumed to travel through undisturbed ambient air within which the spray is uniformly distributed. We assume throughout this study that the droplets are dispersed in the boundary layer and do not directly interact with the ®ber (e.g., through some type of collision and ®lm boiling process). We further assume that the boundary layer remains laminar and that cross¯ows are unimportant. Radiation cooling of small diameter, 1

The short necking region is neglected.

779

Fig. 1. Coordinate system.

high speed ®bers is easily shown to be small relative to forced convective cooling, and it is not included here. 2. Development of boundary layer evaporation model The boundary layer equations in cylindrical coordinates for laminar, steady state, incompressible ¯ow on a continuous moving ®ber in still air are @u @v v ‡ ‡ ˆ0 @x @y a ‡ y @u @u @ 2u 1 @u ru ‡ rv ˆm ‡ 2 @x @y @y a ‡ y @y

u

…1† !

  @T @T k @ @T 1 ‡v ˆ ‡ q_ …a ‡ y † @x @y rcp …a ‡ y † @ y @y rcp e

…2†

…3†

where u is the axial velocity, v is the radial velocity, q_ e is a volumetric heat source due to evaporation (W/m3), y is measured from the ®ber surface, and the coordinate system is further described in Fig. 1. The position y ˆ 0 corresponds to the ®ber surface, and x ˆ 0 to the ®ber bushing plate. The ®ber radius and ®ber velocity are assumed constant over the length of the ®ber.1 All equations are written in terms of y rather than r to simplify the boundary conditions. Eqs. (1) and (2) are subject to the boundary conditions u ˆ U, u 4 0,

v ˆ 0 at y ˆ 0 v40

as y 4 1

…4† …5†

where U is the ®ber speed, and Eq. (3) is subject to the boundary conditions

780

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

Fig. 2. Boundary layer control volume.

T…x, y † ˆ Tf …x†

T…x, y † 4 T1 ,

at y ˆ 0 @T 4 0 as y 4 1 @y

…6†

  u 1 y ˆ 1 ÿ ln 1 ‡ U a a

…7†

u ˆ 0 for yrd…x† U

where Tf …x† is the local ®ber temperature. An additional boundary condition on Eq. (3) is q_e jyˆ0 ˆ 0. This condition arises from the assumption that no drops touch the surface of the ®ber. Following the work of Glauert and Lighthill [3] for axial ¯ow over a long cylinder and the analyses of a continuous cylinder moving through a still ¯uid by Sakiadis and others [4,8], an approximate velocity pro®le of the form

for yRd…x†

…8†

…9†

is assumed, where a is a function of x only, and the spray is assumed to have negligible e€ect on the velocity boundary layer. (The number density of spray is relatively low and the typically boundary layer thicknesses on long thin ®bers are many times greater than typical droplet diameters.) By setting u=U ˆ 0, the boundary layer thickness d, can be de®ned as d ˆ a…ea ÿ 1†. As shown by Bourne and Elliston [5], an integral analysis of Eqs. (1) and (2) allows the relation-

Fig. 3. Fiber control volume.

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

ship of the velocity pro®le parameter, a, to x to be evaluated as 2nx e2a ÿ 1 ‡ Ei…2a † ‡ ln…2a † ‡ g ÿ 2 ˆ Ua2 a

…10†

where g ˆ 0:5572 . . . is Euler's constant and Ei…2a† is the exponential integral. Note that Eq. (10) shows a to be a function of the parameter X ˆ nx=Ua2 . To obtain a corresponding equation for the temperature pro®le in the presence of evaporation, we consider the boundary layer control volume shown in Fig. 2. Neglecting conduction in the x direction, an energy balance yields … d 1 rcp T…x,y †u…x,y †2p…a ‡ y † dy Dx dx 0 …1 @ T ˆ ÿ2pak Dx ‡ …11† q_e 2p…a ‡ y † dy Dx @ y yˆ0 0 where u…x,y† is the boundary layer velocity pro®le, T…x,y† is the boundary layer temperature pro®le and k is the thermal conductivity of the air. Another equation is needed to determine the axial variation in ®ber temperature. For ®bers with diameters less than 200 mm, Papamichael and Miaoulis [25] have shown that the e€ect of axial conduction can be neglected. The analysis of Chida and Katto [7] shows that radial conduction depends upon the parameter B ˆ …krcp †=…kf rf cpf †; for glass ®bers drawn through air, B is of the order of 1.6  10ÿ5, and radial conduction may be ignored. Using a control volume on a section of the ®ber (Fig. 3), energy conservation for the ®ber is described by  d 2 @ T pa rf cpf Tf U ˆ 2pak …12† dx @ y yˆ0 where rf is the ®ber density, cpf is the ®ber speci®c heat, and conduction along and across the ®ber is neglected. De®ning the temperature defects Y…x,y † ˆ T…x,y † ÿ T1 Yf …x† ˆ Tf …x† ÿ T1

and

…13†

a temperature distribution similar to the velocity distribution can be assumed [6]   Y 1 y for yRdT ˆ 1 ÿ ln 1 ‡ …14† Yf b a Y ˆ 0 for yedT Yf where b is a function of x. This form of temperature pro®le is acceptable in so far as the evaporative term

781

in Eq. (3) is smaller than the others and acts to thin the boundary layer over some distance in x, while having a small in¯uence on the local shape in y. Evaporative e€ects will modify the evolution of b with x. Rearranging Eq. (12) and substituting for the assumed temperature pro®le yields arf cpf U dYf ÿkYf ˆ dx ba 2

…15†

which describes the decay of the ®ber temperature. After substituting for the velocity and temperature pro®les and evaluating the integral on the left-hand side of Eq. (11) with the assumption dRdT , the equation becomes " d 1 a2 Yf Ue2a …1 ÿ a ‡ b† dx 4 ab 1 a2 Yf U…2ab ‡ a ‡ 1 ‡ b† ÿ 4 ab ˆ

#

  … ÿk ÿYf 1 dT ‡ q_ …a ‡ y † dy: rcp rcp 0 e b

…16†

Expanding the left-hand side of Eq. (16) leads to an expression in da=dx, dYf =dx and db=dx, which can be simpli®ed by using Eq. (15) to eliminate dYf =dx and by using da 2na2 ˆ dx Ua2 … ÿ e2a ‡ ae2a ‡ a ‡ 1 †

…17†

(from Eq. (10)) to eliminate da=dx and to transform db=dx to db=da. A di€erential equation for b is obtained  db 2b Z ÿ 2a ˆ ‡ e ÿ ae2a ‡ be2a ÿ 2ab ÿ a ÿ 1 ÿ b da aPr a2 Pr  ÿ b 2ae2a ÿ 2a2 e2a ‡ 2abe2a ÿ e2a ÿ be2a ‡ 1 ‡ b ÿ … ÿ e2a ‡ ae2a ‡ a ‡ 1 † a … dT 2b2 ‡ …18† q_ …a ‡ y † dyP Yf akPr 0 e where Pr is the Prandtl number and Z ˆ …rcp †=…rcp †f ; in the absence of evaporative cooling, b ˆ b…X,Pr,Z†. This result is correct as long as the velocity boundary layer d is smaller than the thermal boundary layer dT , which is the normal case for a Prandtl number less than unity. With the addition of spray, however, the thermal boundary layer will actually start to shrink until the thermal boundary layer will be fully contained by the velocity boundary layer. When this happens, the integration across the boundary layer of Eq. (11) is evaluated from 0 to dT ˆ a…eb ÿ 1† rather than 0

782

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

to d ˆ a…ea ÿ 1†. Using the same method as used to develop Eq. (18), the di€erential equation for b becomes db ÿ2ab ˆ da Pr 2



2ab2 ÿ Y kPr 2 f



3  e2a e2a ÿ 1 e2a 1 2 ÿ ÿ ‡ 6 7 4ÿ a2 a a a 5 2 ÿ 2be2b ÿ 2abe2b ‡ 2b e2b ‡ e2b ‡ ae2b ÿ a ÿ 1 3  e2a e2a ÿ 1 e2a 1 ÿ ‡ ÿ 6 7 4ÿ a2 a a a 5 2 2b 2b 2b 2b 2b ÿ 2be ÿ 2abe ‡ 2b e ‡ e ‡ ae ÿ a ÿ 1 … dT q_e …a ‡ y † dy 2

0

 Zÿ ÿ e2b ÿ ae2b ‡ be2b ‡ 2ab ‡ a ‡ 1 ‡ b ‡ Pr 2 3   e2a e2a ÿ 1 e2a 1 2 ÿ ÿ ‡ 6 7 4ÿ 5 a2 a a a  ÿ 2be2b ÿ 2abe2b ‡ 2b2 e2b ‡ e2b ‡ ae2b ÿ a ÿ 1 ÿ  ÿ e2b ‡ be2b ‡ 1 ‡ b b : ‡ ÿ a ÿ 2be2b ÿ 2abe2b ‡ 2b2 e2b ‡ e2b ‡ ae2b ÿ a ÿ 1 The presence of Yf in Eqs. (18) and (19) prevents direct integration of b with a. In addition, the evaporation term will be shown to be a function of Yf . Eqs. (18) and (19) can be used to evaluate the change in the thermal boundary layer due to normal boundary layer growth and due to evaporation within the boundary layer. The value of b can be evaluated as a function of a which in turn can be used to evaluate the ®ber position. Knowledge of the thermal boundary layer at any point on the ®ber can then be used to evaluate the heat transfer from the ®ber to the surroundings.

3. Modeling evaporation Our aim is to evaluate the evaporation integrals in Eqs. (18) and (19). We do this by calculating the total evaporation rate in successive di€erential control volumes of length Dx along the ®ber; axial variations in the evaporation are accounted for in the subsequent numerical integration of the equations in x. The evaporation rate in the increment Dx is proportional to the total droplet surface area and a temperature-dependent evaporation coecient. The total droplet surface area, in turn, depends upon the droplet size distribution and number density. For evaporation, the droplet size distribution is best described using the Sauter mean diameter (SMD), which is the diameter of a drop having

the average surface area to volume ratio of the entire spray cloud. Several approximations are made in modeling the droplet evaporation, the justi®cation for which is discussed in detail in Ref. [26]. The transient warm-up time of each drop presumed short compared to droplet lifetime; the droplets quickly reach the wet-bulb temperature appropriate to the local boundary layer air temperature and humidity. The droplets in the boundary layer are assumed to be fully entrained and to travel at the air velocity corresponding to the radial position of the center of the drop. The droplets in the boundary layer never touch the ®ber surface while ®ber temperature is above the droplet saturation temperature; the solutions obtained apply only to this nonwetting regime. The droplets in the boundary layer are assumed to be uniform at the SMD. A well-established formula exists for the evaporation rate of a single droplet [27]: D2 ˆ D20 ÿ kevap t

…20†

where D is the droplet diameter at time t, D0 is the original droplet diameter, and kevap is an evaporation constant that is dependent on temperature and relative humidity and which can be calculated using mass transfer theory [28]. The coecient kevap was calculated for a temperature range from 300 to 1500 K using both low and high rate mass transfer theory, and the resulting data were ®t to second-order polynomial expressions of the form kevap ˆ A1 T 2 ‡ B1 T ‡ C1

…21†

where T is the local air temperature, which will vary with position y across the thickness of the boundary layer and with distance x along the ®ber. The air is assumed to have zero relative humidity, giving an upper bound on the cooling e€ect. Evaluation of the total mass evaporated, after integration of our equations, shows that axial increases in the relative humidity of air in the boundary layer produce no more than a 4% decrease in the evaporation rate, and generally much less than that. We have, therefore, neglected the e€ect of axial changes in the relative humidity. Within an annular section of the di€erential control volume, the volumetric evaporative heat loss is simply the latent heat of vaporization at droplet temperature, hfg, times the mass evaporation rate per unit volume, m_ 000 (kg/m3 s), evaluated locally. Thus, the evaporative cooling rate of the boundary layer increment can be expressed as an integral through the radial thickness of the boundary layer:

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

… dT 0

q_ e 2p…a ‡ y † dy Dx ˆ

… dT 0

becomes

hfg m_ 000 2p…a ‡ y † dy Dx: …22†

The evaporation rate is the time rate of change of the mass of droplets per unit volume:     d p d p m_ e000 ˆ nrw D3 ˆ u…x,y † nrw D3 dt 6 dx 6   p dn dD3 ˆ u…x,y †rw D3 ‡n …23† 6 dx dx for n the local number density (drops/m3), D the Sauter mean diameter of the spray, and u…x,y† the local air speed. In order to satisfy the condition that no droplets touch the ®ber, a linear droplet density distribution is assumed near the ®ber nˆ

n0 y d

…24†

where n0 is the droplet density outside the boundary layer and d is the local boundary layer thickness.2 The number density, n0, is assumed to be independent of x after the point where spraying is initiated; the spray is also assumed to have a uniform angular distribution about the ®ber. The evaporation rate can now be written with Eq. (8) as    prw n0 y U y dD 000 m_ ˆ U ÿ ln 1 ‡ D2 …25† 2d a a dx where the term dD=dx can be evaluated with Eq. (20) as 1=2 d d 2 Dˆ D0 ÿ kevap t dx dx  ÿ1=2  1 2 dt : ÿ kevap D0 ÿ kevap t ˆ 2 dx

783

…26†

The kinematic derivative dt=dx can be replaced by 1=u…x,y†, and within the di€erential control volume of length Dx, t can be approximated by Dx=u…x,y†. (For typical values of Dx, D20  kevap t.) The dependence of kevap on local temperature in the boundary layer is represented using the polynomial ®t (Eq. (21)). Eq. (26) 2 A linear pro®le is the simplest assumption that satis®es these boundary conditions; however, the decreasing density is also expected because droplets nearer the ®ber will move at higher axial speed, tending to lower particle concentrations. Given that integral solutions are not strongly sensitive to the pro®le shape assumed, and given that more detailed physical information about the number density pro®le is not currently available, we have chosen to use the simplest approximation.

  !ÿ1=2 A1 T 2 ‡ B1 T ‡ C1 Dx d 1 2 D ˆ ÿ D0 ÿ u…x,y † dx 2 ! A1 T 2 ‡ B1 T ‡ C1 : u…x,y †

…27†

Substituting Eqs. (25), (27) and (20) into Eq. (22) and simplifying, we obtain a discretized formulation of the integral … dT 0

… dT

prw hfg n0 y…a ‡ y † 4d…x† h i 2 A1 T…x,y † ‡B1 T…x,y † ‡ C1 " ÿ  #1=2 A1 T…x,y †2 ‡B1 T…x,y † ‡ C1 Dx 2 D0 ÿ dy Dx: u…x,y † q_ e …a ‡ y † dy Dx ˆ ÿ

0

4. Numerical solution If the boundary layer temperature pro®le at any point on the ®ber is known, then the heat transfer from the ®ber can be calculated directly as Q ˆ ÿ2pak ˆ

  @ T ÿYf Dx ˆ ÿ2pak Dx @ y yˆ0 ba

2pkYf Dx: b

…29†

Hence, a knowledge of b can be used to calculate the axial ®ber temperature distribution (cf. Eq. (12)). Eqs. (18) and (19) can be integrated from an initial a value to a ®nal a value to obtain the b value at any position on the ®ber. The initial condition on Eq. (18) is b ˆ 0 at a ˆ 0, but this results in the equation being of an indeterminant form at a ˆ 0. To evaluate b for small values of a, a series expansion can be used in which b is approximated as b ˆ a1 a ‡ a2 a2 ‡ a3 a3 ‡   

…30†

with a1 ˆ

1 …Pr ‡ 2 † 2 Pr

…31†

a2 ˆ

…Pr ÿ 1 †…Pr ‡ 2 † ‡ 4Z 9Pr…Pr ‡ 1 †

…32†

784

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

Fig. 4. Baseline temperature pro®le for 10 mm ®ber at 60 m/s.

ÿ  …Pr ÿ 1 †…Pr ‡ 2 †…3Pr2 ÿ 4Pr ÿ 2 † ‡ 60 3Pr2 ‡ Pr ‡ 2 Z ‡ 360Z2 a3 ˆ 270…Pr ‡ 1 †2 …3Pr ‡ 2 † In these relations, Z ˆ …rcp †=…rf cpf † as before; Z is of the order of 10ÿ3 for glass ®bers in air. Eq. (30) is used to evaluate b for a values from 0 to 0.08. The results from the series expansion are used as initial conditions for Eq. (18). To obtain a solution to the ®ber temperature pro®le along the length of the ®ber, a Fehlberg fourth±®fth order Runge±Kutta method was used to forward integrate Eqs. (18) and (19). The evaporation integral (Eq. (28)) also lacks a closed form solution and must be evaluated numerically. An adaptive Newton±Cotes method was used for the simulation. The integration step length of a was decreased until the ®ber temperature pro®le converged to within 1 K at 1 m. Decreasing the step length more than this would not change the ®ber temperature pro®le by more than 1 K total no matter how much smaller the a increment was made, but decreasing the increment step further greatly increased the number of calculations and the simulation run time. The simulation is designed to allow for initial spray induction to begin at any point x along the ®ber; the

…33†

spray is present from that location onward. Before the spray appears in the boundary layer, a no-spray solution is calculated by setting the spray density to zero. At the initial point of contact with the spray, the bvalue from the no-spray solution is used to determine the temperature pro®le. This pro®le is assumed constant over the small interval from ai to ai‡1 . This pro®le along with the corresponding x values are used to evaluate the evaporation integral. This numerical result is substituted back into the b di€erential equation which is then forward integrated to ®nd bi‡1 . With this value of b, the change in ®ber temperature over Dx can be found using ÿ2pak DT ˆ

@ T Dx @ y yˆ0

pa2 rf cpf U

ˆ

ÿ2kYf Dx a2 brf cpf U

…34†

The ®ber temperature at x ‡ Dx is simply Tx‡Dx ˆ Tx ‡ DT, and this value is then used as input to evaluate the evaporation term over the next interval. At each point, the a and b-values are compared to

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

785

Fig. 5. Temperature pro®les for varying droplet densities (SMD = 70 mm).

determine which di€erential equation for b (Eq. (18) or (19)) should be used. This procedure is followed over the length of the ®ber to solve for the ®ber temperature pro®le. The droplet diameter distribution at the beginning of any interval Dx is assumed uniform and equal to a speci®ed SMD, representing the conditions in the air outside the boundary layer. Variable property e€ects are handled by using a local average ®lm temperature, and polynomial ®ts to tabulated data. The ®lm temperature is changed for each successive increment of the ®ber. This is in contrast to most published results for ®ber cooling which have assigned a single property reference state along the entire length of the ®ber. Validation tests of the numerical model were performed in the limit of no spray. We ran calculations corresponding to the cases that were evaluated numerically and experimentally by Bourne and Dixon [6], with excellent agreement. As noted in Section 1, subsequent studies of the no-spray situation have repeatedly con®rmed the ®ndings in Ref. [6]. 3 The values predicted by the two theories begin to diverge at about 500 K and di€er by more than a factor of two at an air temperature 1500 K. Note that the water temperature is limited to values below 1008C at an ambient pressure of 1 atm.

5. Results The numerical simulation has been run over a wide range of con®gurations, but for comparison purposes a baseline case was chosen: a 10 mm diameter ®ber drawn at 60 m/s through still air at 320 K. The initial ®ber temperature is 1500 K with a droplet density of 5000 drops/cm3 and a representative SMD of 70 mm; these values are typical of the nozzles used in the industrial glass ®ber forming process. The spray was injected into the ®ber boundary layer at a distance x ˆ 1:8 cm from the beginning of the ®ber (corresponding to a ˆ 5:0). The baseline results for the no-spray and spray-cooled temperature pro®le are shown in Fig. 4. The sprayed ®ber cools signi®cantly more rapidly. For example, without spray the ®ber reaches 400 K at x ˆ 1:2 m, whereas with spray that temperature is reached at x ˆ 0:72 m, a 40% reduction in distance. In terms of temperature di€erences, at a ®ber position of x ˆ 1 m the temperature di€erence between the spray cooled and no spray case is about 90 K (163 F). Fiber temperature pro®les were calculated using evaporation coecients evaluated from both low and high rate mass transfer theory. Although the evaporation coecients from the two theories vary signi®cantly at high temperatures,3 the ®ber cooling predictions of the two theories are almost identical. In the region of the boundary layer where the air tem-

786

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

Fig. 6. Temperature pro®les for constant volume of sprayed liquid (9  10ÿ4 m3 water/m3 air). The uppermost curve is for no spray and the lowermost is for an SMD of 10 mm.

peratures are very high and high rate theory is essential, the total cross-sectional area is very small (proportional to …y ‡ a†2 ). The majority of the boundary layer volume is in the region where the temperatures are lower (T between 400 and 500 K) and the evaporation coecients for the high and low rate theories are close. Most of the evaporative cooling takes place in this outer region of the boundary layer, and the evaporative cooling immediately adjacent to the ®ber is very small owing to the small cross-sectional area and lower droplet density. One of the primary control variables in the manufacturing process is the spray condition, speci®cally, the droplet size distribution and number density. The size distribution can be changed by changing nozzle types or by operating the nozzles at a di€erent line pressure. The number density can be varied by changing the number of nozzles, changing the nozzle placement, changing the nozzle type, or changing the operating pressure. In Fig. 5, the droplet density varies from 0 to 10,000 drops/cm3, while the SMD is held constant. As the droplet density decreases, the temperature pro®le approaches the no spray pro®le. Conversely, as number density increases, more liquid surface area is present and evaporative cooling e€ects are stronger. Finer atomization of a given liquid volume produces stronger evaporative e€ects. Fig. 6 shows the e€ect of reducing the SMD of a ®xed volume of water per unit

volume of air; this is equivalent to using successively ®ner atomizers at a ®xed liquid ¯owrate. The spray volume used corresponds to baseline conditions of 5000 drops/cm3 with an SMD of 70 mm. As SMD is decreased, the spray number density is increased to keep the ratio of water volume to total air volume constant. In this ®gure, as SMD decreases from 300 to 10 mm, the cooling of the ®ber becomes progressively more rapid. The smaller the SMD, the more cooling for the same amount of water. Fig. 7 shows the thermal and momentum boundary layer thickness calculated with and without spray. As has been established by previous studies of long thin ®bers, the boundary layer thickness may be more than two orders of magnitude greater than the ®ber diameter. Spray cooling greatly thins the thermal boundary layer, and this causes a signi®cant increase in the convective cooling of the ®ber. The kink in the predicted boundary layer thickness with spray results from the assumption that the spray is instantaneously introduced at x ˆ 1:8 cm. In reality, of course, the spray will require a certain distance to become distributed into the boundary layer. The axial position at which the spray initially enters the boundary layer has an e€ect on the temperature pro®le. Fig. 8 shows the ®ber temperature pro®les for spray entry points ranging from 1.8 to 71 cm below the bushing plate. In terms of the ®ber temperature

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

Fig. 7. Thermal and velocity boundary layer thickness for baseline ®ber conditions.

Fig. 8. Temperature pro®les for varying spray contact points.

787

788

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

Fig. 9. Fiber temperature pro®le for 10 mm ®ber at 90 m/s.

beyond 90 cm from the bushing plate, it makes very little di€erence where the spray enters the boundary layer. This demonstrates that ®nal ®ber temperature is not sensitive to nozzle positioning along the ®ber axis. If the objective is to simply cool the ®ber before the point where the surfactant is applied, then getting spray into the ®ber bundle and around each ®ber is more important than having the spray contact each ®ber at the same x position. If the physical properties or residual stress state of the ®ber require a certain cooling rate between two particular temperatures, then vertical placement of the spray may become critical. As the position of spray introduction is moved farther down the ®ber, the thickness of the boundary layer into which the spray is injected becomes much greater. This in turn means that the cooling e€ect of the spray is greatly increased, since much more spray is contained in a thicker boundary layer. That is the reason for the steeper cooling rates that occur in Fig. 8 when the spray is added far downstream of the bushing plate. Introducing the farther spray downstream also increases the numerical error in the calculations, owing to the much more abrupt change in the boundary layer growth rate immediately following the point of injection. Errors in the vicinity of the injection point a€ect the temperature pro®le along the remaining length of the ®ber. In our calculations, when the spray was introduced very close to the beginning of the ®ber

…x ˆ 1:8 cm), the numerical increment in a was decreased until the ®ber temperature pro®le converged within 1 K. When the spray was introduced farther downstream, limitations in our computation led to errors of roughly 20 K at a distance x ˆ 1 m; these errors contribute to the separation of the spray-cooled temperature pro®les seen in Fig. 8. The ®ber cooling rates with and without spray are a€ected by the ®ber size and drawing speed. Fig. 9 shows the ®ber temperature pro®le for a 10 mm ®ber moving at 90 m/s. Spray conditions in this case are the same as the baseline, with a spray density of 5000 drops/m3 and a SMD of 70 mm. Compared to the 10 mm ®ber at 60 m/s (Fig. 4), the ®ber temperature is reduced less over the same length of ®ber. The opposite e€ect is seen for a ®ber moving more slowly than the baseline condition. To illustrate the cooling of other sizes of ®ber, Fig. 10 shows the ®ber temperature pro®le for a 25 mm ®ber being drawn at 10 m/s. Spray cooling reduces the temperature at x ˆ 1 m from 490 to 370 K. Fig. 11 shows the temperature pro®le for a 5 mm ®ber being drawn at 50 m/s. Spray cooling has much less of an e€ect on this small ®ber. Without spray cooling, the ®ber has already cooled to the ®nal state value long before reaching 1 m. By x ˆ 0:5 m, the ®ber temperature is 320 K with spray cooling and 370 K without.

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

Fig. 10. Fiber temperature pro®le for 25 mm ®ber at 10 m/s.

Fig. 11. Fiber temperature pro®le for 5 mm ®ber at 50 m/s.

789

790

M. Sweetland, J.H. Lienhard V / Int. J. Heat Mass Transfer 43 (2000) 777±790

6. Summary and recommendations A model has been developed to describe the evaporative cooling of spun glass ®bers by water sprays. This model can be used in designing and optimizing the spray processes used in glass ®ber manufacturing. 0The main implications of the model results are as follow. Spray cooling acts to thin the thermal boundary layers surrounding the ®bers and to raise the convective heat removal from the ®bers. The e€ectiveness of spray cooling can be increased by improving the atomization quality to give smaller droplets and by increasing droplet number density near the ®bers. The majority of the cooling takes place in the outer edges of a ®ber's boundary layer. The vertical placement of the spray nozzles within the ®ber bundle is a less critical parameter. The degree of spray cooling varies with ®ber size and speed. Three issues a€ecting the modeling of spray cooling require further study. The ®rst is the turbulent transition of the boundary layer, which is ill-characterized in the existing literature. The second is the spray entrainment and dispersion within the axisymmetric ®ber boundary layer, and, more generally, within the entire bundle of many hundreds of ®bers. The last is the potential for droplet impact at high number density and high transverse droplet injection speeds. The present results are for laminar boundary layers in which droplets are evenly dispersed and travel at the local air speed, and in which droplets do not directly interact with the ®bers. References [1] K.L. Loewenstein, The Manufacturing Technology of Continuous Glass Fibres, 3rd ed, Elsevier, Amsterdam, 1993. [2] R.A. Seban, R. Bond, Skin friction and heat transfer characteristics of a laminar boundary layer on a cylinder in axial incompressible ¯ow, J. Aeronautical Sci 18 (1951) 671±675. [3] M.B. Glauert, M.J. Lighthill, The axisymmetric boundary layer on a long thin cylinder, Proc. R. Soc. London A320 (1955) 188±203. [4] B.C. Sakiadis, Boundary-layer behavior on continuous solid surfaces: III. The boundary layer on a continuous cylindrical surface, AIChE Journal 7 (3) (1961) 467±472. [5] D.E. Bourne, D.G. Elliston, Heat transfer through the axially symmetric boundary layer on a moving circular ®bre, Int. J. Heat Mass Transfer 13 (1970) 583±593. [6] D.E. Bourne, H. Dixon, The cooling of ®bres in the formation process, Int. J. Heat Mass Transfer 24 (1971) 1323±1332. [7] K. Chida, Y. Katto, Conjugate heat transfer of continuously moving surfaces, Int. J. Heat Mass Transfer 19 (1976) 461±470. [8] J. Karnis-, V. Pechoc-, The thermal laminar boundary

[9] [10] [11] [12]

[13] [14] [15] [16]

[17] [18] [19] [20] [21]

[22]

[23]

[24] [25] [26]

[27] [28]

layer on a continuous surface, Int. J. Heat Mass Transfer 21 (1978) 43±47. R.E. Sayles, Approximate solution for the viscous boundary layer on a continuous cylinder, AIChE Journal 36 (12) (1990) 1917±1919. E. Beese, K. Gersten, Skin friction and heat transfer on a circular cylinder moving in a ¯uid at rest, J. Appl. Math. Physics 30 (1979) 117±127. R.G.C. Arridge, K. Prior, Cooling time of silica ®bers, Nature 203 (1964) 386±387. J.V. Alderson, J.B. Caress, R.L. Sager, The cooling rate of glass ®ber in the continuous ®lament process, Laboratory Report No. L.R. 235 of Pilkington Bros., Lathom, Lancashire, 1968. R. Maddison, P.W. McMillan, The cooling rate of glass ®bres, Glass Technology 19 (5) (1978) 127±129. B.H. Kang, J. Yoo, Y. Jaluria, Experimental study of the convective cooling of a heated continuously moving material, J. Heat Transfer 116 (1) (1994) 199±208. E. Richelle, R. Tasse, M.L. Riethmuller, Momentum and thermal boundary layer along a slender cylinder in axial ¯ow, Int. J. Heat Fluid Flow 16 (1995) 99±105. S. Roy Choudhury, Y. Jaluria, Forced convective heat transfer from a continuously moving heated cylindrical rod in materials processing, J. Heat Transfer 116 (3) (1994) 724±734. L.R. Glicksman, The cooling of glass ®bres, Glass Technology 9 (5) (1968) 131±138. A. Ziabicki, H. Kawai (Eds.), High-Speed Fiber Spinning, Wiley, New York, 1993. N. Moussiopoulos, Numerical simulation of spray cooling pond performance, J. Fluids Engr 109 (1987) 179± 185. S.-J. Chen, A.A. Tseng, Spray and jet cooling in steel rolling, Int. J. Heat Fluid Flow 13 (4) (1992) 358±369. W.M. Healy, P.J. Halvorson, J.G. Harley, S.I. AbdelKhalik, A critical heat ¯ux correlation for droplet impact cooling at low Weber numbers and various pressures, Int. J. Heat Mass Transfer 41 (1998) 975±978. S.S. Kachhwaha, P.L. Dhar, S.R. Kale, Experimental studies and numerical simulation of evaporative cooling of air with a water spray Ð I. Horizontal parallel ¯ow, Int. J. Heat Mass Transfer 41 (2) (1998) 447±464. S.S. Kachhwaha, P.L. Dhar, S.R. Kale, Experimental studies and numerical simulation of evaporative cooling of air with a water spray Ð II. Horizontal counter ¯ow, Int. J. Heat Mass Transfer 41 (2) (1998) 465±474. F.P. Buckingham, A. Haji-Sheikh, Cooling of high temperature cylindrical surfaces using a water±air spray, J. Heat Transfer 117 (4) (1995) 1018±1027. H. Papamichael, I.N. Miaoulis, Axial heat conduction e€ect in the cooling of optical ®bres, Glass Technology 32 (3) (1991) 102±108. M. Sweetland, Nozzle analysis and heat transfer model for spray cooling glass ®bers, S.M. thesis, Massachusetts Institute of Technology, Cambridge, Masachusetts, 1998. A.H. Lefebvre, Atomization and Sprays, Hemisphere, New York, 1989. A.F. Mills, Heat and Mass Transfer, Richard D. Irwin, Boston, 1995.